COHERENCE - Cooking: leaderboard
Metric: Exact-match accuracy (%) on the 1,326 cooking recipe documents; 6,161 interleaved image-text documents whose images were removed: the model assigns each candidate image to its placeholder; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 19 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 (High) | 79.19 |
| 2 | Gemini 3.1 Pro (Preview) | 73.45 |
| 3 | Claude Sonnet 4.6 (Thinking) | 73.15 |
| 4 | Qwen 3.5 397B A17B | 71.27 |
| 5 | Doubao-Seed-2.0-Pro-260215 | 63.27 |
| 6 | Qwen 3.5 122B A10B | 61.54 |
| 7 | Kimi K2.5 | 58.97 |
| 8 | GPT-5.2 | 55.28 |
| 9 | Qwen 3.5 35B A3B | 54.15 |
| 10 | Qwen 3 VL 235B A22B | 46.83 |
| 11 | Qwen 3.5 4B | 43.14 |
| 12 | Step3 VL 10B | 42.84 |
| 13 | GLM-4.6V | 40.27 |
| 14 | Qwen 3 VL 8B Instruct | 31.3 |
| 15 | Qwen 3 VL 4B Instruct | 26.24 |
Interactive version: theaggregate.ai/benchmark?slug=coherence-cooking · How It Works · Data refreshed daily, snapshot 2026-10-07.